Genetic Variation Across the Historical Range of the Wild Turkey

Home , Junglefowl, Ocellated turkey, Partridge

Molecular Ecology (2002) 11, 643–657

BlackwellGenetic Science Ltd variation across the historical range of the wild turkey (Meleagris gallopavo)

K. E. MOCK,*‡ T. C. THEIMER,* O. E. RHODES Jr.,† D. L. GREENBERG* and P. KEIM* *Department of Biological Sciences, Northern Arizona University, Flagstaff AZ 86011–5640, USA, †Department of Forestry and Natural Resources, 1159 Forestry Building, Purdue University, West Lafayette, Indiana 47907, USA

Abstract Genetic differences within and among naturally occurring populations of wild turkeys (Meleagris gallopavo) were characterized across five subspecies’ historical ranges using amplified fragment length polymorphism (AFLP) analysis, microsatellite loci and mitochondrial control region sequencing. Current subspecific designations based on morphological traits were generally supported by these analyses, with the exception of the eastern (M. g. silvestris) and Florida (M. g. osceola) subspecies, which consistently formed a single unit. The Gould’s subspecies was both the most genetically divergent and the least genetically diverse of the subspecies. These genetic patterns were consistent with current and historical patterns of habitat continuity. Merriam’s populations showed a positive association between genetic and geographical distance, Rio Grande populations showed a weaker association and the eastern populations showed none, suggesting differing demographic forces at work in these subspecies. We recommend managing turkeys to maintain subspecies integrity, while recognizing the importance of maintaining regional population structure that may reflect important adaptive variation. Keywords: AFLP, control region, Meleagris gallopavo, microsatellite, phylogeography, wild turkey Received 15 August 2001; revision received 5 December 2001; accepted 5 December 2001

or subspecies ranges. This approach threatens to disrupt Introduction the historical patterns of genetic diversity and gene flow The wild turkey is indigenous to North America, and had and potentially could lead to increased homogenization of an historical range that included much of the eastern and subspecies and the loss of locally adapted gene complexes portions of the southwestern United States, northern Mexico (Dobzhansky 1940; Templeton 1986; Avise 1992; Dickerman and southeastern Canada. This historical distribution was & Hubbard 1994). In other areas, habitat degradation and apparently limited by amount of snowfall to the north and fragmentation may lead to increased isolation of regional the lack of adequate water and roost trees to the west. By populations, resulting in a loss of local genetic diversity the early 1900s, the original range of the wild turkey had (Lacy 1987; Leberg 1991; Leberg et al. 1994). Thus, the decreased significantly due to habitat alterations and historical patterns of genetic diversity and gene flow need over-harvesting. Since that time, wild turkeys have been to be described for wild turkey before they are lost (Leberg re-introduced to some areas and translocated to others, 1991). particularly in the midwestern and northeastern USA Six subspecies of the wild turkey have been recognized: (Mosby 1949; Mosby 1975; Williams 1981; Kennamer & the eastern (Meleagis gallopavo silvestris), Florida (M. g. osceola), Kennamer 1996). Programmes to introduce turkeys into Rio Grande (M. g. intermedia), Merriam’s (M. g. merriami), new, unoccupied habitat, or to re-introduce them into Gould’s (M. g. mexicana), and south Mexican (M. g. gallopavo) historical habitat, often have not considered historical species (Stangel et al. 1992). The eastern subspecies inhabits woodlands ranging from the hardwood forests of the northeastern USA to the oak-hickory forests of the midwestern USA and Correspondence: Dr K. E. Mock, ‡Present address: Department of Fisheries and Wildlife, 5210 Old Main Hill, Utah State Univer- the pine-oak forests of the southeastern USA. Historically, sity, Logan, UT 84322–5210, USA. Fax: 1 435 797 1871; E-mail: its range was continuous with that of the Florida subspecies [email protected] to the south and the Rio Grande subspecies to the west

Fig. 1 Wild turkey sample population locations and historical subspecies boundaries suggested by Schorger (1966). Population location description and sample sizes are provided in Table 1.

(Schorger 1966; Williams 1981; Stangel et al. 1992) (Fig. 1). fragment length polymorphism (AFLP) loci as well micro- The Merriam’s and Gould’s subspecies occupy ponderosa satellite analysis and mitochondrial control region sequences. pine and pine-oak woodlands of the southwestern USA Using these data sets, we investigated (i) the putative his- and northern Mexico. Their ranges are currently isolated torical patterns of genetic diversity across the species’ from those of the other subspecies and from each other by range using relict populations, (ii) the genetic validity of tracts of unsuitable desert and grassland habitat. However, current subspecific designations, and (iii) patterns of gene the ranges of Merriam’s, Gould’s and Rio Grande subspecies flow across geographical distance within subspecies. may have been continuous during the expansion of forest habitat in the late Holocene. Materials and methods Current turkey subspecies designations have been based on body size differences and plumage variation, such as Sample acquisition and DNA extraction wing-barring patterns, body colour and extent and colour of edging on the tail coverts and rectrices. However, these Because of the long history of human manipulation of wild morphological characteristics have not been quantitatively turkey populations, we focused on those populations that analysed, and may not be reliable indicators of genetic most probably represented ‘relicts’ of historical patterns of divergence (Schorger 1966; Stangel et al. 1992). Preliminary genetic diversity. We defined these populations as those genetic analyses demonstrated statistically significant (i) located within the historical geographical range of the differences among all except the eastern and Florida species; (ii) having no recorded introductions; and (iii) having subspecies, with the Gould’s subspecies being the most no known extirpations or severe declines that might have genetically divergent (Mock et al. 2001). However, these encouraged introductions from other areas or created a conclusions were based on a limited data set, used a single genetic bottleneck. We chose populations that were relatively marker system, and averaged allele frequencies within evenly distributed across the historical range of each subspecies rather than assessing genetic patterns among subspecies to reduce biases due to geographical proximity. populations. In this study we expand these earlier genetic Because of their persistence, most of these populations analyses (Mock et al. 2001) to include additional amplified have been used extensively as translocation sources during

© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 643–657 RANGEWIDE GENETIC ANALYSIS OF WILD TURKEY 645 the restoration efforts of the past several decades. Although obtained from the Yucatan Peninsula, Mexico. All samples we feel that these populations are the best available were obtained at voluntary hunter check stations from 1995 representation of the historical patterns of genetic diversity to 1999, with the exception of the samples from Colorado and divergence in wild turkey, we recognize the possibility and Sonora, and a portion of the samples from Missouri that undocumented translocations may have altered these and South Carolina, which were collected by winter trapping. patterns to some extent, particularly in the eastern USA. Tissue samples collected from hunter-killed birds were We sampled a total of 24 populations, representing 379 pulverized in the field using razor blades and 0.2–2.0 g was individuals. These included six eastern populations (total immediately suspended in 5 mL lysis buffer (Longmire n = 105), three Florida populations (total n = 46), seven Rio et al. 1988). Blood samples from trapped birds were obtained Grande populations (total n = 77), five Merriam’s populations by jugular venepuncture, and were either collected directly (total n = 110), and three Gould’s populations (total n = 41) into heparinized tubes and kept on ice or 0.5 mL was (Table 1, Fig. 1). Additionally, four tissue samples from collected into a sterile tube containing 5 mL of lysis buffer. hunter-killed ocellated turkeys (Meleagris ocellata) were Blood samples collected into heparinized tubes were

Table 1 Sample population locations, sample sizes and numbers of samples from each population used for various analyses

Population Location Samples (n) Mito. (n) Msat (n) AFLP (n)

Eastern ESW Scotch WMA*, Alabama 20 10 20 20 EWW Waterhorn WMA, S. Carolina 20 10 19 20 EOM Ozark Mts, Missouri 18 9 18 18 EWV various locations, W. Virginia 19 10 19 19 EBW Black Warrior WMA, Alabama 18 10 — 18 ECL Camp Lejeune, N. Carolina 10 7 10 10 Totals 105 56 86 105 Florida FAA Avon Air Force Base, Florida 21 10 21 21 FTL Three Lakes WMA, Florida 9 9 9 9 FBC Big Cypress WMA, Florida 16 10 16 15 Totals 46 29 46 45 Rio Grande RDW James E. Daughtery WMA, Texas 15 12 — 15 REN Encino, Texas 9 9 9 9 RTC Terrell County, Texas 5 5 5 5 RGH Gene Howe WMA, Texas 10 10 10 10 RKC Kleberg County, Texas 11 11 11 11 RNL private ranch, Nuevo Leon, Mexico 7 6 7 7 RKW Kerr WMA, Texas 20 9 20 20 Totals 77 62 62 77 Merriam’s MCS Colorado Springs, Colorado 20 10 20 20 MSP Spanish Peaks, Colorado 40 20 20 40 MSL Stoneman Lake, Colorado 16 15 16 16 MWM White Mountains, Arizona 22 21 23 22 MSM Ruidoso, New Mexico 12 7 12 12 Totals 110 73 91 110 Gould’s GYS Yecora, Sonora, Mexico 21 10 21 21 GNS northern Sierra Madre, Chihuahua, Mexico 10 6 10 10 GCS central Sierra Madre, Chihuahua, Mexico 10 9 10 10 Totals 41 25 41 41 O Ocellated Turkey 4 4 — —

Mito. = mitochondrial control region sequencing, Msat = microsatellite analysis. Population designations correspond with locations depicted in Fig. 1. The first letter of each population abbreviation refers to the subspecies designation (E = eastern, F = Florida, R = Rio Grande, M = Merriam’s, G = Gould’s). *WMA = wildlife management area.

Table 2 Primers used for control region ′ Primer Sequence 5 -chicken position sequencing in wild turkeys

LND6-2 5′-aga aaa atc aca aaa taa gtc a 16683 (+ strand) H125-2 5′-act tgc atg tat atg tct agc a 1351 (– strand) L16750 (Fumihito et al. 1994) 5′-agg act acg gct tga aaa gc 16731 (+ strand) HD4 (Edwards 1993) 5′-ggt acc atc ttg gca tct tc 1280 (– strand) NAU313 5′-gcc acc tgt gaa gaa gcc 681 (– strand) NAU185 5′-acg gct tga aaa gcc att gtt gt 16737 (+ strand)

The positions of the 5′ nucleotides of these primers with reference to the chicken (Gallus gallus) mitochondrial genome (Dejardins & Morais 1990; GenBank NC001323) are provided in the last column. suspended in lysis buffer (0.5 mL sample aliquots in 5 mL Rapid PCR Purification tubes and resuspended in TE buffer lysis buffer) upon arrival in the laboratory. A 1.5-mL aliquot (10 mm Tris, 1 mm ethylenediaminetetraacetic acid). of each of the sample/buffer mixtures was combined with Sequencing reactions were performed on these amplicons 500 µg proteinase K and 100 µL 0.1 m dithiothreitol and using the ABI BigDye™ kit with the forward primers digested overnight in a 55 °C water bath. DNA extraction used to produce the amplicon, and sequencing products was performed using a salt–chloroform technique with were precipitated using the manufacturer’s ethanol– isopropanol precipitation (Mullenbach et al. 1989). The quality sodium acetate procedure. An ABI 377 sequencer (Applied and quantity of DNA was assessed by electrophoresis in Biosystems Inc., Foster City, CA) was used to generate 0.7% agarose gels stained with ethidium bromide and by sequence chromatograms that were manually proofread comparison to molecular weight and size standards (λ and aligned using ABI sequence navigator™ software. HindIII digest and 100-bp ladder). Only samples yielding These sequences were aligned with other galliform predominantly high quality (> 20 kb) DNA were included sequences and mutations were mapped relative to previ- in the study. ously described conserved areas within the control region (Dejardins & Morais 1990; Quinn & Wilson 1993; Fumihito et al. 1994; Fumihito et al. 1995; Randi & Lucchini 1998). Mitochondrial control region sequencing and data Nuclear copies of mitochondrial genes can confound analysis analyses of mitochondrial sequences (Sorenson & Quinn We sequenced a 438-bp fragment in domain I of the 1998). However, the confirmation of sequence results with mitochondrial control region, putatively the most variable alternate primer pairs, the consistency of results with portion of the control region in vertebrates (Moritz et al. different tissue sources, the absence of spurious amplicons 1987; Randi & Lucchini 1998). A summary of the primers from blood samples, and the alignment of control region used for this purpose is provided in Table 2. Control region sequences with other galliforms all suggest that the ampli- amplicons were obtained for 249 individual birds using the cons were mitochondrial in origin. primers LND6-2 and H125-2, and for eight birds (MSL1, The haplotype and nucleotide diversities (π) for each GYS2, EWW1, EWW4, EWW12, EOM4, EOM6 and FBC1) subspecies group were calculated using dna-sp software using the primers L16750 and HD4. Replicate amplicons (Rozas & Rozas 1997). A neighbour-joining tree of mitotypes were obtained and sequenced as a quality control measure was created based on the proportion of sequence differ- for a total of 32 individual birds, or 13% of the samples, ences among mitotypes using Phylogenetic Analysis Using using alternative primer sets. These replicates included seven Parsimony (paup) software (Swofford 1998), assigning the amplicons obtained using primers LND6-2 and NAU313, ocellated turkey sequences to the outgroup. Average eight amplicons obtained using primers NAU185 and HD4, genetic distances of individuals between each pairwise and 18 amplicons obtained using primers L16750 and HD4. combination of populations, calculated as the proportion Polymerase chain reaction (PCR) amplification reactions of nucleotide differences, were obtained using mega 2.0 containing 200 µm dNTPs, 2 mm MgCl, 1 × PCR buffer, 0.2 µm (Kumar et al. 2001). This matrix was used to construct a primers, 1 U Taq polymerase, and 50 ng DNA in a total neighbour-joining tree of populations using ntsys software µ volume of 50 L were run for 35 cycles with an annealing (Rohlf 1993). Mitochondrial genetic structure (FST) among temperature of 55 °C. Amplicon quantity and quality were subspecies was assessed using both the Analysis of Molecular assessed on 0.7% agarose mini-gels with ethidium bromide Variation (amova) approach (Weir & Cockerham 1984; λ Φ using size (100 bp ladder) and concentration ( HindIII digest) Excoffier et al. 1992; Weir 1996; ST) and the conventional standards. Amplicons producing a single, well-defined band approach (θ) with arlequin software (Schneider et al. 2000). of approximately 1300 bp were purified using Concert™ For the amova analysis, a distance matrix of pairwise

© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 643–657 RANGEWIDE GENETIC ANALYSIS OF WILD TURKEY 647 differences (Nei & Li 1979) was constructed using individual software (Miller 1997). Standard deviations of θ were mitotypes. For the calculation of θ, mitotypes were pooled calculated by jackknifing over loci and 95% confidence according to neighbour-joining results to reduce diversity. intervals for θ were generated by bootstrapping across loci Φ θ The probability of the observed or greater values of both ST (Miller 1997). Pairwise -values between subspecies were and θ was assessed using 1000 permutations of the data. calculated using arlequin software (Schneider et al. 2000). The significance of these pairwise values was assessed using 1023 permutations of the data among populations. Microsatellite analysis Genetic diversity was assessed using microsatellite data Turkeys were genotyped at five microsatellite loci [TUM6 by calculating Nei’s (1978) unbiased gene diversity (D) (U79372), TUM12 (U79356), TUM17 (U79312), TUM23 across all loci within each subspecies and by counting the (U79332), and TUM50 (U79306)], originally characterized total number of alleles represented in each subspecies in domestic turkeys (Huang et al. 1999) and later optimized across all five loci. In order to assess the significance of for eastern wild turkeys (Shen 1999). Microsatellite loci pairwise subspecific differences in D, pairwise differences were amplified as follows: genomic DNA (30 ng) was used were calculated on a locus-specific basis and a 95% con- as a template in a 10-µL reaction mixture containing 0.2 mm fidence interval for the overall difference was generated each dATP, dTTP, dGTP and dCTP; 2 pmol each locus- using 1000 bootstrap replicates. A 95% confidence interval specific primer, one of which contained an M13 universal that did not include zero was taken as evidence of signi- primer tag at the 5′ end; 0.1 pmol M13 universal primer ficant difference in D. Because sample size can affect with IRD florescent dye tag; 1 × reaction buffer containing the number of alleles observed, the lowest observable

1.5 mm MgCl2, 50 mm KCL, 10 mm Tris–HCL pH 9.0, 0.1% frequency was calculated for the Gould’s subspecies, which Triton X-100; 1.0 U Taq DNA Polymerase (Promega). was represented by the smallest number of samples (41). PCR reactions were carried out in 96-well microtitre plates Hence, alleles at frequencies below 0.025 were not counted and sealed with a microseal A sealing film (MJ Research) in the total number of alleles. under the following conditions: a 5-min denaturation step In order to investigate the possibility of a recent bottle- at 95 °C; four cycles of 95 °C for 45 s, 68 °C (decreasing by neck in each subspecies, we used the program bottleneck 2 °C at each cycle) for 5 min; two cycles of 72 °C for 1 min, (Cornuet & Luikart 1996) to calculate the one-tailed prob- 95 °C for 45 s, 58 °C (decreasing by 2 °C per cycle) for two ability of excess observed heterozygosity (gene diversity) minutes, 72 °C for one minute; 31 cycles of 95 °C for 45 s, at these microsatellite loci relative to the expected gene 54 °C for 2 min, 72 °C for 1 min; terminal extension step of diversity under the two-phased model of mutation, using 72 °C for 10 min. Products were electrophoresed on Long the Wilcoxon sign-rank test (Luikart et al. 1998). If a popu- Ranger Single polyacrylamide gels (FMC Bioproducts) using lation has undergone a recent bottleneck, it is expected that a Licor DNA Analyser 4200. Alleles were assigned using the number of alleles in that population will decline more RFLPScan (Scanalytics), with commercial molecular weight rapidly than the gene diversity, leading to excess gene markers and duplicate individuals as scoring standards. diversity relative to the expectation based on allele number Analysis of these microsatellite loci, each consisting of (Luikart et al. 1998). Additionally, we used bottleneck dinucleotide repeats, was performed on 249 samples (Table 1). (Cornuet & Luikart 1996) to investigate the shape of the Microsatellite data were analysed in a hierarchical manner allele frequency distribution, which is expected to deviate using Tools for Population Genetic Analysis (TFPGA; from the l-shaped distribution found in stable populations if Miller 1997). Hardy–Weinberg genotypic proportions were there has been a recent bottleneck (Luikart & Cornuet 1997). tested with TFPGA using a Markov chain approach to esti- Microsatellite data were also used to characterize regional mate the probability of the observed genotypic proportions patterns of geographical versus genetic distances. Pairwise (Haldane 1954; Guo & Thompson 1992; Miller 1997). θ-values were calculated for all pairs of populations within upgma cluster analysis of populations based on a matrix the eastern, Rio Grande and Merriam’s subspecies and of Nei’s (1978) unbiased genetic distances was performed plotted against the corresponding pairwise geographical using ntsys (Rohlf 1993) software, and the strength of the distances to provide a visual comparison of patterns among major internal nodes was assessed by bootstrapping over subspecies, following Hutchison & Templeton (1999). Mantel loci with 1000 pseudoreplicates using tfpga software (Miller testing was used to assess the strength and significance of 1997). In order to compare among- versus within-subspecies the correlation between the pairwise matrices of θ and dis- genetic distances, we calculated the average (± SE) pairwise tance (Mantel 1967) using ntsys software (Rohlf 1993). Nei’s (1978) unbiased distances between all pairs of subspecies and populations within each subspecies. The AFLP analysis strength of genetic structuring at both the population and subspecific levels was assessed with Weir & Cockerham’s AFLP marker profiles were generated for 378 individuals θ (1984) theta ( ), an estimator of Wright’s FST, using tfpga (Table 1) using procedures described by Vos et al. (1995)

© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 643–657 648 K. E. MOCK ET AL. and modified by Travis et al. (1996) and Busch et al. (2000). Table 3 Mitotypes found in populations of wild turkey (**Gen- We used a total of eight EcoRI/MseI primer combinations Bank alignment accession numbers AF486875s–AF487121) in the second selective amplification (selective nucleotides Mitotype Populations ACG/AGC, ACG/AGG, ACG/ACC, AGG/ACG, ACC/ ACG, AGG/ACC, ATC/AGG, and ATC/ACC) to generate A MCS(8), MRM(7), MSL(7), MSM(3), MSP(7), markers ranging in size from 90 to 500 bp. Polymorphic MWM(11) markers were selected in the absence of sample labels, and B MSM(1), MSP(1) only if they could be scored unambiguously across all C MSM(2) samples. Five per cent of the samples were completely D MCS(1), MSL(2), MWM(1) E MWM(3) replicated to assess the overall error rate associated with F MWM(3) the analysis. The error rate was calculated as the number of G MRM(1), MSP(1) loci with conflicting scores in replicated samples divided H MSL(5), MSM(1) by the total number of loci scored. Because AFLP analysis I MWM(2) produces dominant markers, the assumption of Hardy– J RGH(2) Weinberg genotypic proportions must be made in order to K RDW(1), RKW(1), RTC(1), REN(1), RKC(3), RKW(4), calculate allele frequencies for statistical analysis (Mueller MSL(1) L RDW(3) & Wolfenbarger 1999). Allele frequencies were estimated M GNS(5), GCS(9), GYS(10) based on the frequency of the homozygous recessive allele N EWW(5), FAA(2), FBC(5), FTL(4), RDW(2), RGH(3) using the Taylor expansion approach of Lynch & Milligan O FBC(4) (1994). Cluster analysis, analysis of population and P FAA(2), FBC(1), ESW(5), EOM(6), EBW(3), MCS(1), subspecies genetic structure, gene diversity analysis, and MRM(1), REN(1), RGH(3), RKW(1), RTC(3), RKC(1) analysis of regional patterns of gene flow were performed Q EWV(2) for AFLP data in the same manner as described above for R FTL(1), FAA(3) S ECL(4) microsatellite data. In addition, the per cent polymorphic T REN(1), RGH(1), RKC(2), RNL(3) loci (P) was calculated for subspecies as an independent U EBW(3) measure of genetic diversity. The 95% confidence intervals V EOM(1), ESW(2), EWV(1), RDW(2) were generated for P by scoring each locus as a one W EBW(4) (polymorphic according to the 95% criterion) or a zero X EOM(2) (monomorphic) for each subspecies and performing 1000 Y FTL(2), EWV(2), FAA(2) bootstrap replicates. Z ECL(1), EWV(4), RKC(2), EWW(1) AA MRM(1), MSP(1), MWM(1) AB EWW(2) Results AC RDW(2) AD RDW(1), REN(5), RGH(2), RKC(3), RKW(2), RNL(3) Mitochondrial sequencing and analysis AE ECL(2), EWW(2) gnc15 GNC(1) Alignment of mitochondrial control region sequences from rdw11 RDW(1) 245 individual birds across all five subspecies (Tables 1 and 3, rkw3 RKW(1) esw3 ESW(1) excluding the ocellated turkey) yielded 42 mitotypes, with esw1 ESW(1) 31 polymorphic nucleotide sites, 23 of which were phylo- eom5 EOM(1) genetically informative (Table 3). All informative nucleotide ewv6 EWV(1) variation was in the form of transitions, and most mito- ftl1 FTL(1) types differed by only one to two mutations. Mutations ftl9 FTL(1) among wild turkeys occurred primarily in clusters between faa23 FAA(1) known conserved areas in domain I of the galliform control ren3 REN(1) region (Randi & Lucchini 1998). The neighbour-joining Numbers in parentheses following the population codes refer to tree describing phylogenetic relationships among mitotypes the number of individuals within each population. suggested a lack of diagnostic structure among the eastern, Florida and Rio Grande mitotypes, but strong clustering of the Merriam’s mitotypes (Fig. 2). The Gould’s populations II) of the control region. The neighbour-joining tree based were fixed for a single mitotype, with the exception of on differences among individuals (Fig. 3) followed the a closely related mitotype found in a single individual. same general pattern seen in the haplotype tree (Fig. 2), One of the mutations found uniquely in Gould’s mitotypes suggesting that the Gould’s subspecies was the most and in one Florida mitotype (represented by a single indi- divergent, followed by the Merriam’s and Rio Grande vidual) was located in the central conserved region (domain subspecies.

Fig. 2 Neighbour-joining dendrogram of wild turkey mitotypes based on control region sequence data. Mitotypes are designated by letter, number of individuals represented, and subspecies: eastern subspecies (E), Florida subspecies (F), Rio Grande subspecies (R), Merriam’s subspecies (M), Gould’s subspecies (G). Numbers along the right margin designate mitotype clusters used for calculation of F-statistics.

Eight mitotype clusters were assigned, based on the Microsatellite analysis neighbour-joining mitotype tree (Fig. 2), excluding the θ ocellated turkey samples, in order to estimate FST using . Microsatellite markers provide multiallelic, co-dominant A very high degree of mitotype structuring among subspe- genetic data to contrast AFLP analysis and control region Φ θ cies was detected ( ST = 0.613, = 0.546 (P (both) < 0.0001). sequencing. Amplification of microsatellite loci was suc- Pairwise θ-values among subspecies (Table 4) suggested cessful on all but 28 of the 1655 total (individual by locus) that the eastern and Florida subspecies were the least reactions. The number of alleles detected per locus was distinct, and the θ calculated for this pair of subspecies was highly variable: TUM6 (two alleles, 145–147 bp), TUM12 (12 the only nonsignificant comparison (P = 0.009) using the alleles, 191–223 bp), TUM17 (14 alleles, 161–189 bp), Bonferroni-adjusted significance level of 0.005. TUM23 (12 alleles, 144–166 bp), and TUM50 (38 alleles, Mitotype diversity (h) and nucleotide diversity (π) were 113–160 bp). Genotypic frequencies were significantly similar among subspecies with the exception of the Gould’s different (P < 0.05) from Hardy–Weinberg expectations in subspecies, which was less diverse by an order of magni- eight of the 105, or 7.6%, of the population by locus data tude according to both measures (Table 5). The remaining sets for which there was more than one segregating allele. subspecies were characterized by high h and low π. The The deviations were found in the following populations eastern and Rio Grande subspecies showed the greatest (and loci): FBC (TUM12), RDW (TUM50), RKC (TUM12), diversity according to both measures. RKW (TUM50), EOM (TUM 12), MWM (TUM17), EWV

Table 4 Pairwise θ-values among subspecies of wild turkey using mitochondrial sequences (above diagonal), AFLP data (above diagonal, italics), and microsatellite data (below diagonal)

Eastern Florida Gould’s Merriam’s Rio Grande

Eastern — 0.120* 0.687 0.679 0.252 0.035 0.320 0.176 0.106 Florida 0.003a — 0.654 0.658 0.178 0.336 0.181 0.126 Gould’s 0.010 0.010 — 0.888 0.554 0.287 0.255 Merriam’s 0.010 0.010 0.014 — 0.559 0.151 Rio Grande 0.003 0.003 0.009 0.008 —

*not significant (P ≥ 0.005, with Bonferroni correction).

Table 5 Haplotype (h) and nucleotide (π) diversities (Nei 1987) for wild turkey subspecies based on mitochondrial control region sequencing

Subspecies h (SD) π (SD)

Eastern 0.895 (0.03) 0.00451 (0.0004) Florida 0.813 (0.05) 0.00402 (0.0005) Rio Grande 0.868 (0.02) 0.00547 (0.0003) Merriam’s 0.643 (0.06) 0.00378 (0.0008) Gould’s 0.080 (0.07) 0.00018 (0.0002)

level (θ = 0.173, SD = 0.05) than at the subspecies level (θ = 0.113, SD = 0.03), and none of the 95% confidence intervals for these values included zero (Fig. 5). Values of θ for pairwise subspecies (Table 4) suggested that the least structuring was between eastern and Florida, eastern and Rio Grande, and Rio Grande and Florida Fig. 3. Neighbour-joining dendrogram of wild turkey popu- subspecies, while the greatest degree of structuring was lations based on a matrix of the average proportion of control region nucleotide differences. Population codes are described in between the Merriam’s and Gould’s subspecies. All θ Table 1. pairwise -values were significant (P < 0.005, with a Bonferroni correction for pairwise comparisons) except for the eastern and Florida subspecies (P = 0.0185). Values (TUM 12), GYS (TUM 17). Thus, there was no apparent of θ for populations within subspecies suggested that the concentration of these deviations within a particular Merriam’s populations were the most highly structured, population, subspecies, or locus. followed by Rio Grande populations (Fig. 5). There was upgma cluster analysis of populations (Fig. 4a) was a negative correlation (r = 0.81) between the number of consistent with current subspecific designations with the alleles at a locus and the θ-value calculated for that locus. exception of the eastern and Florida subspecies, which This correlation was nonsignificant (P = 0.10) with only formed a single cluster. Bootstrap support for the major five loci. cluster topology, however, was weak. Average genetic dis- According to the analysis of paired, microsatellite tances between pairs of populations within subspecies allele-specific differences in D, the eastern subspecies was were consistently and significantly less than distances significantly more diverse than any of the other species, among subspecies except the eastern and Florida sub- and the Rio Grande subspecies was more diverse than species (Table 6). The average genetic distance between either the Gould’s or Merriam’s subspecies. Other differ- eastern and Florida populations was within a standard error ences in D among pairs of subspecies were not significant. of the average genetic distance between eastern popula- The total number of microsatellite alleles found across tions. Genetic structuring was stronger at the population all five loci were as follows: Merriam’s (22), Gould’s (24),

CS Fig. 4 upgma cluster analysis dendro- SP SM Merriam's grams based on Nei’s (1978) distance WM calculated from (a) five microsatellite loci 0.44 SL and (b) 83 AFLP loci in wild turkey a. NL EN populations representing five subspecies. 0.54 KC Rio Numbers at interior nodes represent the GH Grande KW proportion of 1000 bootstrap replicates that TC clustered the same taxa. CS YS Gould's NS SW OM WV Eastern/ AA Florida TL 0.95 CL BC WW

BW SW WV AA Eastern/ WW Florida 1.0 TL OM BC 0.91 CL b. NL EN KC Rio DW Grande 1.0 KW TC GH WM SL MS Merriam's SP SM YS GC Gould's NS

0.15 0.10 0.05 0

Nei’s (1978) Genetic Distance

Table 6 Matrix of average pairwise Nei’s (1978) unbiased genetic distances (± SE) within and among subspecies using AFLP (A, above the diagonal) and microsatellite data (M, below the diagonal). Intrasubspecific contrasts are italicized

Eastern Florida Rio Grande Merriam’s Gould’s

Eastern 0.0312 (A) (0.003) 0.0311 (0.002) 0.0644 (0.003) 0.0943 (0.002) 0.1451 (0.004) 0.1269 (M) (0.027) Florida 0.1710 (0.019) 0.0263 (A) (0.004) 0.0638 (0.004) 0.0885 (0.003) 0.1425 (0.004) 0.1049 (M) (0.025) Rio Grande 0.3490 (0.038) 0.4929 (0.044) 0.0356 (A) (0.003) 0.0900 (0.003) 0.1209 (0.005) 0.1362 (M) (0.022) Merriam’s 0.6255 (0.058) 0.8119 (0.048) 0.2780 (0.020) 0.0537 (A) (0.006) 0.1318 (0.005) 0.1590 (M) (0.026) Gould’s 0.4409 (0.065) 0.4409 (0.065) 0.3265 (0.022) 0.3367 (0.027) 0.0197 (A) (0.002) 0.0344 (M) (0.017)

Florida (31), Rio Grande (32), and eastern (34). The Florida eastern (P = 0.109), or Gould’s (P = 0.406) subspecies. None subspecies did show evidence of a recent bottleneck of the subspecies showed a deviation from the l-shaped according to the one-tailed Wilcoxon test for gene diversity distribution of allele frequencies expected for stable pop- excess (P = 0.03125). The results of this test were not signi- ulations (Luikart et al. 1998). Of the three subspecies tested ficant for the Merriam’s (P = 0.969), Rio Grande (P = 0.500), for a pattern of isolation by distance using microsatellite

AFLP analysis AFLP analysis yielded 83 polymorphic loci that could be scored for all samples. The technique error rate was estimated at 2.5%, which is comparable to the error rates found in other AFLP studies on natural populations (Busch et al. 2000). Genetic clustering (Fig. 4b) followed existing subspecific boundaries, with the exception of the eastern and Florida subspecies, supporting the microsatellite results (Fig. 4a). Bootstrap support for the major cluster topology was greater than 90%. Consistent with the microsatellite results, average genetic distances between pairs of populations within subspecies were significantly less than distances among subspecies with Fig. 5 Values of θ and 95% confidence intervals for AFLP and the exception of the eastern and Florida subspecies microsatellite data among subspecies, all populations, and popu- (Table 6). lations within subspecies. As with microsatellite data, genetic structuring with AFLP data was found to be greater at the population level data, only the Merriam’s subspecies showed a significant (θ = 0.294, SD = 0.02) than at the subspecies level (θ = 0.175, correlation (r = 0.656) between the pairwise matrices of θ SD = 0.02), and at neither level did the 95% confid- and geographical distance with Mantel testing (P = 0.043) ence intervals of θ include zero. Within subspecies, the (Fig. 6). Merriam’s populations showed the highest levels of genetic structuring (θ) (Fig. 5). AFLP data yielded higher θ-values

θ AFLP data: Microsatellite data: Fig. 6 Pairwise population comparisons of theta ( ) M. g. merriami M. g. merriami versus geographical distance in three subspecies of 0.30 0.30 r = 0.856 r = 0.656 p = 0.043 wild turkey using AFLP and microsatellite data. 0.25 p = 0.012 0.25 Values of r and P provided in each graph were 0.20 0.20 calculated from a Mantel test comparing matrices of θ 0.15 0.15 and geographical distance. Theta 0.10 Theta 0.10 0.05 0.05 0.00 0.00 0 500 1000 1500 0 500 1000 1500 Kilometres Kilometres

AFLP data: Microsatellite data: M. g. silvestris M. g. silvestris 0.30 r = 0.165 p = 0.233 0.30 r = –0.485 p = 0.074 0.25 0.25 0.20 0.20 0.15 0.15 Theta Theta 0.10 0.10 0.05 0.05 0.00 0.00 0 500 1000 1500 0 500 1000 1500 Kilometres Kilometres

AFLP data: M . g. intermedia Microsatellite data: 0.30 r = 0.343 p = 0.183 0.30 M. g. intermedia 0.25 0.25 r = 0.295 p = 0.217 0.20 0.20 0.15 0.15

Theta 0.10 0.10 Theta 0.05 0.05 0.00 0.00 0 500 1000 1500 –0.05 0 500 1000 1500 Kilometres Kilometres

Table 7 Allele-specific differences in D between all pairs of subspecies expressed as 95% confidence intervals for AFLP (above diagonal) and microsatellite (below diagonal) data

Eastern Florida Rio Grande Merriam’s Gould’s

Eastern P (AFLP) = 0.50–0.70 0.008–0.034 –0.049–0.011 –0.051–0.028 –0.039–0.054 Florida 0.006–0.133 P (AFLP) = 0.39–0.61 –0.062–0.001 –0.021–0.065 –0.045–0.050 Rio Grande 0.020–0.060 –0.108–0.052 P (AFLP) = 0.37–0.58 –0.042–0.031 –0.071–0.018 Merriam’s 0.076–0.287 –0.255–0.003 –0.254 to –0.023 P (AFLP) = 0.46–0.67 –0.066–0.024 Gould’s 0.045–0.344 –0.312–0.073 –0.308 to –0.001 –0.284–0.248 P (AFLP) = 0.54 –0.73

Confidence intervals in bold did not include zero. Per cent polymorphic loci (P) for each subspecies, also expressed as 95% confidence intervals, are provided along the diagonal.

than microsatellite data at both the species and subspecies may also have obscured historical differences among these levels (Fig. 5). The eastern subspecies had a significantly two groups, but our sample populations were chosen higher value of D than the Florida subspecies, but all other specifically to minimize this possibility. The Rio Grande pairwise differences in D among subspecies were non- subspecies historically occupied habitat continuous with significant (Table 7). The 95% confidence intervals for P that of the eastern/Florida subspecies as well (Schorger were overlapping for all subspecies. 1966; Aldrich 1967). It is possible that these subspecies may As with the microsatellite data, only the Merriam’s sub- now appear to be genetically distinct only because species showed a significant correlation (r = 0.856) between intermediate populations occupying habitat near the pairwise matrices of θ and geographical distance with historical range boundaries were extirpated in the early Mantel testing (P = 0.012) (Fig. 6). 1900s. However, there was no indication of an east–west cline in genetic differences across eastern populations, as might be expected in such a situation. Discussion Genetic differentiation of the Merriam’s and Gould’s subspecies is most likely the result of their relatively long Genetic differentiation and subspecific designation geographical isolation from other subspecies by unsuitable Our data generally support current subspecies designations desert habitats. Lack of palaeontological evidence of tur- for wild turkey, with the exception of the eastern and Florida keys in the current range of the Merriam’s subspecies has subspecies, which together form a genetically distinct group. led to the hypothesis that this subspecies was introduced The geographical differentiation of turkey populations into its current range less than 1500 years ago by Pueblo along subspecies lines is consistent with the geographical Indian cultures (Hargrave 1970; Rea 1980; McKusick 1986; potential for gene flow and isolation that has historically Breitburg 1988). According to these hypotheses, Merriam’s existed between current subspecies ranges. Continuous turkey was derived from either eastern or Gould’s popu- habitat has linked populations of eastern and Florida lations. Our data do not support a strong link between wild turkeys both currently and historically (Schorger Merriam’s and either of these two groups. An alternative 1966; Aldrich 1967), and gene flow between the two has hypothesis (Shaw & Mollohan 1992) argues that Merriam’s apparently been great enough to prevent major genetic turkey was isolated from Rio Grande and/or Gould’s tur- differentiation detectable with neutral markers. Sub- keys when suitable woodland habitat contracted and was specific plumage characteristics are intermediate in replaced by desert habitats at the end of the Pleistocene and turkeys from the areas where the ranges of these two early Holocene. The mitochondrial data presented here putative subspecies meet in the southeastern United States suggests that the Merriam’s subspecies has been more (Aldrich 1967), a finding consistent with the hypothesis recently associated with the Rio Grande subspecies than of significant gene flow between these two areas. In north- with the Gould’s subspecies. ern Florida and southern Georgia, Remington (1968) Although both nuclear techniques supported the same described a ‘suture zone’ of putative secondary contact in subspecific groupings, they differed with respect to rela- a variety of terrestrial taxa between endemic forms in tionships among the subspecies. The pattern of clustering Florida and continental forms. We found no evidence in the mitochondrial data (Figs 2 and 3) resembled that found of endemic forms of wild turkey in Florida, although such with AFLP data, suggesting that the Gould’s subspecies a pattern could have been eliminated by extensive was the most divergent, followed by the Merriam’s sub- historical and present-day gene flow. Undocumented species. Given the relatively small number of microsatellite translocations between eastern and Florida subspecies loci and the weak bootstrap support for the microsatellite

© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 643–657 654 K. E. MOCK ET AL. groupings, the relationships suggested by AFLP analysis combination with the cluster analysis results, suggests appear the more robust of the two. The relationships that the currently designated subspecies do reflect groups suggested by the microsatellite analysis, with the eastern and of populations that share a long evolutionary history, with Florida subspecies being the most divergent, would also be the exception of the eastern and Florida subspecies. Values difficult to explain based on our understanding of historical of θ estimated using AFLPs were greater than those habitat connectivity. The development of additional micro- estimated using microsatellite data both within and among satellite loci for use in wild turkeys would allow a more subspecies (Fig. 5). This may be an artefact of the high direct comparison of these techniques. One limitation level of polymorphism in microsatellite loci (Hedrick 1999; of AFLP data is the need to assume Hardy–Weinberg equi- Balloux et al. 2000), a conclusion supported by a nega- librium in order to estimate allele frequencies. Because tive correlation between the number of microsatellite we also used microsatellites to characterize the same alleles at a locus and the locus-specific value of θ. The sample set, we were able to assess the validity of this comparative magnitude of θ within subspecies suggests assumption, at least with respect to the microsatellite loci that there is greater genetic differentiation among the analysed. Our results suggest that the number of popula- Merriam’s and Rio Grande populations than among tion by locus data sets that were significantly (P < 0.05) populations within the other subspecies. This may reflect different from Hardy–Weinberg expectations (7.6%) was the less continuous nature of the habitat historically only slightly greater than what would be expected by occupied by these subspecies. chance alone (5%). Therefore, the assumption of Hardy– Weinberg equilibrium in our sample populations seems Geographic versus genetic distance among subspecies reasonable. In the absence of human intervention, dispersal in wild turkeys is most likely by juvenile (predominantly female) Genetic diversity within subspecies birds moving out of the home range of their natal flock into While population-level genetic diversity measures adjacent areas over relatively small geographical distances suggested that no one subspecies was consistently more (Ellis & Lewis 1967; Eaton et al. 1976; Exum et al. 1985). As diverse than the others (Tables 4 and 6), the eastern a result, wild turkey populations may correspond subspecies tended to be among the most diverse and the relatively well to predictions based on two-dimensional Gould’s subspecies tended to be the least diverse, stepping-stone models of isolation by distance (Kimura particularly with respect to mitochondrial diversity. The 1953; Malècot 1955; Kimura & Weiss 1964). These models lack of mitochondrial diversity in Gould’s turkeys could be predict a positive relationship between geographical due to a population bottleneck (Grant & Bowen 1998), but and genetic distance in populations approximating the approach of Luikart et al. (1998) for detecting recent demographic equilibrium (Malecot 1955; Slatkin 1993; bottlenecks based on relative rates of heterozygosity and Rousset 1997; Hutchison & Templeton 1999). The absence allele number losses failed to provide evidence of such a of a strong isolation by distance pattern may have a variety process in the Gould’s subspecies. This lack of diversity of interpretations (Rousset 1997; Hutchison & Templeton could also be due to a persistently low population size in 1999). this subspecies. Unfortunately, information on population The eastern wild turkey and Rio Grande populations we trends in this subspecies is not available. analysed did not show significant isolation by distance The remaining subspecies showed a general pattern of effects, and both showed less structuring among popula- high mitotype diversity and low nucleotide diversity among tions than the Merriam’s populations (Figs 5 and 6). The mitotypes. Grant & Bowen (1998) suggest that such a pat- Rio Grande pattern suggests some effect of distance on tern could be attributed to a recent expansion from a period proximal populations, although the relationship is not sig- of low population sizes, where the low sequence diversity nificant. It is possible that a pattern of isolation by distance among mitotypes is due to a recent coalescence, and the would fail to be detected in these subspecies if distances high mitotype diversity is due to the retention of mutations between sampled populations were too large (Rousset 1997). in an expanding population. In wild turkeys, this may be a The eastern and Rio Grande subspecies were sampled over reflection of rapid expansion following the retreat of glacial somewhat greater geographical distances than was the maxima at the end of the Pleistocene. Merriam’s subspecies. However, even at distances of under 800 km, the pattern of isolation by distance was more apparent among Merriam’s populations than among east- Hierarchical genetic structuring ern or Rio Grande populations, and the values of θ among Genetic structuring (θ) across the species complex was pairwise populations in both the Rio Grande and eastern evident at both the population and subspecies levels subspecies tended to be lower than those among Merriam’s (Fig. 5). The magnitude of θ at the subspecies level, in populations. The apparent absence of a significant isolation

© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 643–657 RANGEWIDE GENETIC ANALYSIS OF WILD TURKEY 655 by distance pattern among the Rio Grande and eastern Management implications populations could also be interpreted as the result of post- Pleistocene colonization into a relatively continuous habi- Based on extant populations that best represent the tat in these subspecies. High levels of gene flow among historical pattern of genetic differentiation, our analyses populations could result in a weaker relationship between support current subspecies designations, except for the genetic and geographical distance relative to the Merriam’s eastern and Florida subspecies. These two subspecies subspecies. This interpretation is consistent with our appear to be a single unit genetically. The remaining understanding of subspecies-specific differences in histor- subspecies appear genetically distinct and may exhibit ical habitat continuity: the eastern and Rio Grande subspe- different demographic structures. Therefore, managers cies occupy habitat that is far more continuous than that of should avoid translocations that would threaten currently the Merriam’s subspecies. It is also possible that this pattern recognized subspecies population integrity. could reflect the influence of widespread reintroduction Here we have used analyses of putatively neutral programmes during the last century, producing a pattern markers to delineate groups of populations that share a consistent with recent range expansion from a common common evolutionary history. However, it is possible that source population (Hutchison & Templeton 1999). These important adaptive traits will not follow patterns in neutral reintroduction programmes were particularly prevalent in variation (Endler 1986; Knapp & Rice 1998; Crandall et al. the eastern subspecies but also occurred in the Rio Grande 2000). Adaptive variation may develop despite high levels subspecies. However, Leberg et al. (1994) have demon- of gene flow, or selective pressures may prevent adaptive strated that the genetic effects of re-introductions on neigh- divergence despite neutral divergence. In general, we are bouring populations may be relatively slow to develop. lacking data on heritable, adaptive variation in wild turkeys. Additionally, our study focused on those populations with In order to address adaptive variation directly, managers the best record of having remained historically strong and should maintain records of translocation outcomes and which tended to serve as source populations for reintro- attempt to correlate these with regional ecological data. duction programmes. Until data on the ecological equivalency of subspecies is In contrast to eastern and Rio Grande populations, available, conservative management should focus at the pairwise θ-values among pairs of Merriam’s popula- level of regional populations while respecting subspecies tions showed a significantly positive relationship with boundaries. geographical distance with both nuclear marker systems The Gould’s subspecies appears to be the most divergent (Fig. 6). This pattern is consistent with expectations for of the five subspecies, but is the least studied in terms of populations that have experienced restricted geographical habitat requirements and availability, demographic history dispersal for long enough to have approached regional and current status. Addressing these questions should become demographic equilibrium. Because suitable habitat for a high priority for managers concerned with the conserva- Merriam’s turkey often forms an archipelago of high- tion of species-wide genetic diversity in wild turkeys. elevation habitat patches separated by unsuitable desert environments, Merriam’s populations are likely to exper- ience lower levels of interpopulational gene flow than Acknowledgements eastern and Rio Grande populations, resulting in a more This study was supported by the National Wild Turkey Federation strongly positive relationship between geographical dis- and the Arizona Game and Fish Department Heritage Fund. We tance and genetic distance (Rousset 1997; Hutchison & are grateful to the following for sample collection assistance: Arizona Templeton 1999). However, further reduction in the Game and Fish Department (B. F. Wakeling, R. Engel-Wilson), limited habitat connectivity that exists among Merriam’s Colorado Department of Natural Resources, Division of Wildlife populations or an increase in translocations could alter (R. W. Hoffman), Texas Parks and Wildlife Department (M. J. Peterson, the relative effects of genetic drift or gene flow, resulting in J. R. Fugate, M. D. Hobson, W. E. Armstrong, J. P. Hughes, W. B. Russ), Alabama Game and Fish Division (S. W. Barnett), West Virginia a gradual loss of this isolation by distance pattern in the Division of Natural Resources (W. A. Lesser, C. I. Taylor), South future (Leberg 1991; Waples 1998; Hutchison & Templeton Carolina Department of Natural Resources (D. P. Baumann), Camp 1999). Lejeune Fish and Wildlife Division (K. A. Lombardo), Avon Park Given the many assumptions underlying isolation by Air Force Range (P. B. Walsh, J. M. Lee), Florida Game and Fresh- distance models and the multitude of factors that can affect water Fish Commission (L. S. Perrin, H. G. Steelman, S. L. Glass), US population structure, hypotheses like those outlined above Fish and Wildlife Service National Forensics Laboratory (S. R. Fain), must remain tentative (Slatkin 1993). Nevertheless, these Real Turkeys (L. E. Williams, Jr.), Missouri Department of Con- servation (L. D. VanGilder, P. C. Freeman, D. A. Hasenbeck), the analyses indicate that the demographic factors that have United States Embassy in Mexico City (A. Narvaez), S. D. Schemnitz, resulted in the current population structure have been R. C. Dujay, G. A. Wright, R. H. Crawford, and A. M. Sada. We different for the Merriam’s versus the Rio Grande and thank Christy J. W. Watson, Emily K. Latch, Baoshan Shen and eastern populations. Christen L. Williams for their assistance with the laboratory portion

© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 643–657 656 K. E. MOCK ET AL. of this research. We also acknowledge H. G. Shaw for his con- Grant WS, Bowen BW (1998) Shallow population histories in deep tinuous inspiration and encouragement, and Joseph Busch for evolutionary lineages of marine fishes: insights from sardines his analytical advice. and anchovies and lessons for conservation. Journal of Heredity, 89, 415–425. Guo SW, Thompson EA (1992) Performing the exact test of References Hardy–Weinberg proportion for multiple alleles. Biometrics, 48, 361–372. Aldrich JW (1967) Historical background. In: The Wild Turkey and Haldane JBS (1954) An exact test for randomness of mating. Journal its Management (ed. Hewitt OH), pp. 3–16. The Wildlife Society, of Genetics, 52, 631–635. Washington DC. Hargrave LL (1970) Feathers from a sand dune cave: a basket- Avise JC (1992) Molecular population structure and the biogeo- maker cave near Navajo Mountain, Utah. Flagstaff: Museum of graphic history of a regional fauna: a case history with lessons Arizona Technical Series, 9, 1–53. for conservation biology. Oikos, 63, 62–76. Hedrick PW (1999) Highly variable loci and their interpretation in Balloux F, Brunner H, Lugon-Moulin N, Hausser J, Goudet J (2000) evolution and conservation. Evolution, 53, 313–318. Microsatellites can be misleading: an empirical and simulations Huang HB, Song YQ, Hdei M etal. (1999) Development and char- study. Evolution, 54, 1414–1422. acterization of genetic mapping rersources for the turkey Breitburg E (1988) Prehistoric new world turkey domestication: (Meleagris gallopavo). Journal of Heredity, 90, 240–242. origins, development, and consequences. PhD Thesis, Southern Hutchison DW, Templeton AR (1999) Correlation of pairwise Illinois University, Carbondale. genetic and geographic distance measures: inferring the relative Busch JD, Miller MP, Paxton EH, Sogge MK, Keim P (2000) Genetic influences of gene flow and drift on the distribution of genetic variation in the endangered southwestern willow flycatcher. variability. Evolution, 53, 1989–1914. Auk, 117, 586–595. Kennamer JE, Kennamer MC (1996) Status and distribution of the Cornuet JM, Luikart G (1996) Description and power analysis of wild turkey in 1994. Proceedings of the National Wild Turkey two tests for detecting recent population bottlenecks from allele Symposium, 7, 203–211. frequency data. Genetics, 144, 2001–2014. Kimura M (1953) ‘Stepping-stone’ model of population. Annual Crandall KA, Binida-Emonds ORP, Mace GM, Wayne RK (2000) Report of the National Institute of Genetics, 3, 62–63. Considering evolutionary processes in conservation biology. Kimura M, Weiss GH (1964) The stepping stone model of popula- Trends in Ecology and Evolution, 15, 290–295. tion structure and the decrease of genetic correlation with Dejardins P, Morais R (1990) Sequence and gene organization of distance. Genetics, 49, 561–576. the chicken mitochondrial genome: a novel gene order in higher Knapp EE, Rice KJ (1998) Comparisons of isozymes and quantitative vertebrates. Journal of Molecular Biology, 212, 599–634. traits for evaluating patterns of genetic variation in purple Dickerman RW, Hubbard JP (1994) An extinct subspecies of sharp- needlegrass (Nassella pulchra). Conservation Biology, 12, 1031–1041. tailed grouse from New Mexico. Western Birds, 25, 128–136. Kumar S, Tamura K, Jakobsen IB, Nei M (2002) MEGA2: Molecular Dobzhansky T (1940) Speciation as a stage in evolutionary diver- Evolutionary Genetics Analysis software. Bioinformatics, 17, 1244– gence. American Naturalist, 74, 312–321. 1245. Eaton SW, Evans JW, Glidden JW, Penrod BD (1976) Annual range Lacy RC (1987) Loss of genetic diversity from managed popu- of wild turkeys in southwestern New York. New York Fish and lations: interacting effects of drift, mutation, immigration, Game Journal, 23, 20–33. selection, and population subdivision. Conservation Biology, 1, Edwards SV (1993) Mitochondrial gene genealogy and gene flow 143–158. among island and mainland populations of a sedentary song- Leberg PL (1991) Influence of fragmentation and bottlenecks on bird, the grey-crowned babbler. Evolution, 47, 1118–1137. genetic divergence of wild turkey populations. Conservation Ellis JE, Lewis JB (1967) Mobility and annual range of wild turkeys Biology, 5, 522–530. in Missouri. Journal of Wildlife Management, 31, 568–581. Leberg PL, Stangel PW, Hillestad HO, Marchinton RL, Smith MH Endler JA (1986) Natural Selection in the Wild. Princeton University (1994) Genetic structure of reintroduced wild turkey and Press, Princeton, NJ. white-tailed deer populations. Journal of Wildlife Management, Excoffier L, Smouse P, Quattro J (1992) Analysis of molecular 58, 698–711. variance inferred from metric distances among DNA haplotypes: Longmire JL, Lewis AK, Brown NC etal. (1988) Isolation and application to human mitochondrial DNA restriction data. molecular characterization of a highly centromeric tandem Genetics, 131, 479–491. repeat in the family Falconidae. Genomics, 2, 14–24. Exum JH, McGlincy JA, Speake DW, Buckner JL, Stanley FM (1985) Luikart G, Cornuet JM (1997) Empirical evaluation of a test Evidence against dependence upon surface water by turkey for identifying recently bottlenecked populations from allele hens and poults in southern Alabama. Proceedings of the National frequency data. Conservation Biology, 12, 228–237. Wild Turkey Symposium, 5, 83–89. Luikart G, Sherwin WB, Steele BM, Allendorf FW (1998) Usefulness Fumihito A, Miyake T, Sumi S-I etal. (1994) One subspecies of the of molecular markers for detecting population bottlenecks via red junglefowl (Gallus gallus gallus) suffices as the matriarchic monitoring genetic change. Molecular Ecology, 7, 963–974. ancestor of all domestic breeds. Proceedings of the National Lynch SM, Milligan BG (1994) Analysis of population genetic Academy of Science USA, 91, 12505–12509. structure with RAPD markers. Molecular Ecology, 3, 91–99. Fumihito A, Miyake T, Takada M, Ohno S, Kondo N (1995) Malècot G (1955) The decrease of relationship with distance. The genetic link between the Chinese bamboo partridge Cold Springs Harbor Symposium on Quantitative Biology, 20, 52– (Bambusicola thoracica) and the chicken and junglefowls of the 53. genus Gallus. Proceedings of the National Academy of Science USA, Mantel N (1967) The detection of disease clustering and a generalized 92, 11053–11056. regression approach. Cancer Research, 27, 209–220.

McKusick CR (1986) Southwest Indian turkeys — prehistory and Schneider S, Roessli D, Excoffier L (2000) Arlequin, Version 2000: comparative osteology. Southwest Bird Laboratory, Globe, a Software for Population Genetics Data Analysis. Genetics AZ. and Biometry Laboratory, University of Geneva, Geneva, Miller MP (1997) Tools for population genetic analysis (TFPGA), Switzerland. 1.3: a Windows program for the analysis of allozyme and Schorger AW (1966) The Wild Turkey: its History and Domestication. molecular population genetic data. Computer software distrib- University of Oklahoma Press, Norman, OK. uted by author. Shaw HG, Mollohan C (1992) Merriam’s Turkey. In: The Wild Mock KM, Theimer TC, Greenberg DL, Keim P (2001) Conserva- Turkey: Biology and Management (56 Dickson JG), pp. 331–349. tion of genetic diversity within and among subspecies of wild Stackpole Books, Harrisburg, PA. turkey. Proceedings of the National Wild Turkey Symposium, 8, in Shen B (1999) Development of genetic markers for cost effective press. surveys in wild turkeys. MS Thesis, Purdue University. Moritz C, Dowling TE, Brown WM (1987) Evolution of animal Slatkin M (1993) Isolation by distance in equilibrium and non- mitochondrial DNA: relevance for population biology and equilibrium populations. Evolution, 47, 264–279. systematics. Annual Review of Ecology and Systematics, 18, Sorenson MD, Quinn TW (1998) Numts: a challenge for avian 269–292. systematics and populations biology. Auk, 115, 214–221. Mosby HS (1949) The present status and future outlook of the east- Stangel PW, Leberg PL, Smith JI (1992) Systematics and popula- ern and Florida wild turkeys. Transactions of the North American tion genetics. In: The Wild Turkey: Biology and Management (ed. Wildlife Conference, 14, 346–358. Dickson JG), pp. 18–28. Stackpole Books, Harrisburg, PA. Mosby HS (1975) The status of the wild turkey in 1974. Proceedings Swofford DL (1998) Phylogenetic Analysis Using Parsimony, Version of the National Wild Turkey Symposium, 3, 22–26. 4.0b8. Smithsonian Institution, Washington, DC. Mueller UG, Wolfenbarger LL (1999) AFLP genotyping and finger- Templeton AR (1986) Coadaptation and outbreeding depression. printing. Trends in Evolution and Ecology, 14, 389–394. In: Conservation Biology: the Science of Scarcity and Diversity Mullenbach R, Lagoda PL, Welter C (1989) An efficient salt- (ed. Soule ME), pp. 105–116. Sinauer Publishers, Sunderland, chloroform extraction of DNA from blood and tissues. Trends in MA. Genetics, 5, 391. Travis SE, Maschinski J, Keim P (1996) An analysis of genetic Nei M (1978) Estimation of average heterozygosity and genetic variation in Asrtagalus cremnophylax var cremnophylax, a criti- distance from small numbers of individuals. Genetics, 89, cally endangered plant, using AFLP markers. Molecular Ecology, 583–590. 5, 735–745. Nei M (1987) Molecular Evolutionary Genetics. Columbia University Vos P, Hogers R, Bleeker M etal. (1995) AFLP: a new technique for Press, New York. DNA fingerprinting. Nucleic Acids Research, 23, 4407–4414. Nei M, Li WH (1979) Mathematical model for studying genetic Waples R (1998) Separating the wheat from the chaff: patterns of variation in terms of restriction endonucleases. Proceedings of the genetic differentiation in high gene flow species. Journal of National Academy of Science, USA, 76, 5269–5273. Heredity, 89, 438–450. Quinn TW, Wilson AC (1993) Sequence evolution in and around Weir BS (1996) Genetic Data Analysis II: Methods for Discrete Popula- the mitochondrial control region in birds. Journal of Molecular tion Genetic Data. Sinauer Associates, Inc, Sunderland, MA. Evolution, 37, 417–425. Weir BS, Cockerham CC (1984) Estimating F-statistics for the ana- Randi E, Lucchini V (1998) Organization and evolution of the lysis of population structure. Evolution, 38, 1358–1370. mitochondrial DNA control region in the avian genus Alectoris. Williams LE (1981) The Book of the Wild Turkey. Winchester Press, Journal of Molecular Evolution, 47, 449–462. Tulsa, Oklahoma. Rea AM (1980) Late Pleistocene and Holocene turkeys in the southwest. Contributions in Science, Natural History Museum of Los Angeles County, 330, 210–223. This work represents a portion of Karen Mock’s PhD dissertation Remington CL (1968) Suture–zones of hybrid interaction between describing genetic diversity among relict and reintroduced recently joined biotas. Evolutionary Biology, 2, 321–438. populations of wild turkey. Raw data are available upon request. Rohlf FJ (1993) NTSYS-Pc: Numerical Taxonomy and Multivariate Analyses reported here were conducted in the laboratory of Analysis System, Version 1.8. Exeter Software, Setauket, NY. Dr Paul Keim at Northern Arizona University. His labor- Rousset F (1997) Genetic differentiation and estimation of gene atory analyses genetic diversity in plants, animals and bacteria. flow from F-statistics under isolation by distance. Genetics, 145, Karen Mock is currently an assistant professor in the Fisheries 1219–1228. and Wildlife Department at Utah State University, where she has Rozas J, Rozas R (1997) Dnasp, Version 2.0: a novel software established a genetics laboratory focusing on phylogeographical package for extensive molecular population genetics analysis. patterns in a variety of species. Computer Application Biosciences, 13, 307–311.